]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
19 years agoInstead of a rather useless generation number, use a sample number to
phk [Sat, 19 Mar 2005 12:55:46 +0000 (12:55 +0000)]
Instead of a rather useless generation number, use a sample number to
keep track of things.

19 years agoBump documentation date for last commit before ru reminds me.
simon [Sat, 19 Mar 2005 08:57:00 +0000 (08:57 +0000)]
Bump documentation date for last commit before ru reminds me.

19 years agoAdd nForce4 to the list of supported audio devices.
simon [Sat, 19 Mar 2005 08:54:06 +0000 (08:54 +0000)]
Add nForce4 to the list of supported audio devices.

MFC after: 1 week

19 years agoUse subr_unit
phk [Sat, 19 Mar 2005 08:22:36 +0000 (08:22 +0000)]
Use subr_unit

19 years agoSleeping is not allowed in uma->fini
phk [Sat, 19 Mar 2005 08:22:13 +0000 (08:22 +0000)]
Sleeping is not allowed in uma->fini

19 years agoDon't free the SIM object twice.
scottl [Sat, 19 Mar 2005 06:32:57 +0000 (06:32 +0000)]
Don't free the SIM object twice.

Submitted by: Coverity Prevent analysis tool

19 years agoHandle failures better in the passthrough bus creation code.
scottl [Sat, 19 Mar 2005 06:29:32 +0000 (06:29 +0000)]
Handle failures better in the passthrough bus creation code.

Submitted by: Coverity Prevent analysis tool

19 years agocheck copyin return value
sam [Sat, 19 Mar 2005 04:34:23 +0000 (04:34 +0000)]
check copyin return value

Noticed by: Coverity Prevent analysis tool

19 years agoAdd device ID for Planet IA 128P-STD card.
murray [Sat, 19 Mar 2005 03:00:22 +0000 (03:00 +0000)]
Add device ID for Planet IA 128P-STD card.

PR: kern/71923
Submitted by: Vassily Kiryanov <admin@assist.berezniki.ru>
MFC after: 1 week

19 years agoTypo.
murray [Sat, 19 Mar 2005 02:44:41 +0000 (02:44 +0000)]
Typo.

PR: conf/77178
Submitted by: Andras Jako <jako.andras@eik.bme.hu>
MFC after: 1 week

19 years agoAdd Central European and Slovak keymaps to sysinstall.
murray [Sat, 19 Mar 2005 02:28:02 +0000 (02:28 +0000)]
Add Central European and Slovak keymaps to sysinstall.

PR: misc/78701
MFC after: 1 week

19 years agoWhen disarming a watchdog by using an interval of WD_TO_NEVER a non-zero
marius [Sat, 19 Mar 2005 01:46:37 +0000 (01:46 +0000)]
When disarming a watchdog by using an interval of WD_TO_NEVER a non-zero
return value of the ioctl doesn't indicate that the command has failed
so don't let watchdog(8) return an error in this case.

MFC after: 3 days

19 years agoOptimize putc routine to write 2 ints instead of 8 chars to uncached
grehan [Sat, 19 Mar 2005 01:43:29 +0000 (01:43 +0000)]
Optimize putc routine to write 2 ints instead of 8 chars to uncached
framebuffer memory. Speeds up system time of large ascii file cat
by 4x.

Inspired by:  Linux scrolling so damned fast.

19 years agoAdd device id for nForce 4 audio controller.
murray [Sat, 19 Mar 2005 01:39:13 +0000 (01:39 +0000)]
Add device id for nForce 4 audio controller.

PR: kern/78482
Submitted by: Markus Niemistö <markus.niemisto@iki.fi>
MFC after: 1 week

19 years ago- Don't call rpcclnt_realign() if we don't have any mbufs to realign.
das [Sat, 19 Mar 2005 01:16:25 +0000 (01:16 +0000)]
- Don't call rpcclnt_realign() if we don't have any mbufs to realign.
- Remove a bogus and unneeded null pointer check.

Found by: Coverity Prevent analysis tool
Approved by: alfred

19 years agoAdd a driver for the 'clock-board' device (the clock board is an
marius [Sat, 19 Mar 2005 01:04:48 +0000 (01:04 +0000)]
Add a driver for the 'clock-board' device (the clock board is an
inevitable component in Sun Exx00 machines and provides serial ports,
NVRAM and TOD amongst others which are handled by uart(4) and eeprom(4)
respectively). This driver currently only prints out information about
the chassis on attach and allows to blink the 'Cycling' LED (which is
duplicated on the front panel) of the clock board just like fhc(4) does
for the other boards. The device name for the LED is /dev/led/clockboard.

Obtained from: OpenBSD
Tested by: joerg

19 years ago- Add a device interface method for bus_get_resource_list() and use
marius [Sat, 19 Mar 2005 00:50:28 +0000 (00:50 +0000)]
- Add a device interface method for bus_get_resource_list() and use
  bus_generic_rl_release_resource() for the bus_release_resource() method
  instead of a local copy.
- Correctly handle pass-through allocations in fhc_alloc_resource().
- In case the board model can't be determined just print "unknown model"
  so the physical slot number is reported in any case.
- Add support for blinking the 'Cycling' LED of boards on a fhc(4) hanging
  of off the nexus (i.e. all boards except the clock board) via led(4).
  All boards have at least 3 controllable status LEDs, 'Power', 'Failure'
  and 'Cycling'. While the 'Cycling' LED is suitable for signaling from
  the OS the others are better off being controlled by the firmware.
  The device name for the 'Cycling' LED of each board is /dev/led/boardX
  where X is the physical slot number of the board. [1]

Obtained from: OpenBSD [1]
Tested by: joerg [1]

19 years agoEnrich with some register descriptions and additional register macros.
marius [Sat, 19 Mar 2005 00:48:34 +0000 (00:48 +0000)]
Enrich with some register descriptions and additional register macros.

Obtained from: OpenBSD

19 years ago- Add a device interface method for bus_get_resource_list() and use
marius [Sat, 19 Mar 2005 00:47:02 +0000 (00:47 +0000)]
- Add a device interface method for bus_get_resource_list() and use
  bus_generic_rl_release_resource() for the bus_release_resource() method
  instead of a local copy.
- Correctly handle pass-through allocations in central_alloc_resource().

19 years agoRemove dead code.
das [Fri, 18 Mar 2005 22:33:10 +0000 (22:33 +0000)]
Remove dead code.

Found by: Coverity Prevent analysis tool

19 years agoDon't brelse(bp) if bp is null. Also, eliminate some redundancy
das [Fri, 18 Mar 2005 21:23:32 +0000 (21:23 +0000)]
Don't brelse(bp) if bp is null.  Also, eliminate some redundancy
and dead code.

Found by: Coverity Prevent analysis tool

19 years agoMove the swap_zone == NULL check earlier (i.e. before we dereference
das [Fri, 18 Mar 2005 21:22:48 +0000 (21:22 +0000)]
Move the swap_zone == NULL check earlier (i.e. before we dereference
the pointer.)

Found by: Coverity Prevent analysis tool

19 years agoAdd missing cases for PT_SYSCALL.
das [Fri, 18 Mar 2005 21:22:28 +0000 (21:22 +0000)]
Add missing cases for PT_SYSCALL.

Found by: Coverity Prevent analysis tool

19 years agoIn order to print out the dump dates correctly, the date and ddate fields
imp [Fri, 18 Mar 2005 21:06:54 +0000 (21:06 +0000)]
In order to print out the dump dates correctly, the date and ddate fields
also need to be convereted for old tapes for records of type TAPE.

19 years agoTypo.
phk [Fri, 18 Mar 2005 20:21:46 +0000 (20:21 +0000)]
Typo.

Submitted by: Hiroharu Tamaru <tamaru@myn.rcast.u-tokyo.ac.jp>

19 years agoSync usage and man page with reality. There's no '-c' command line
imp [Fri, 18 Mar 2005 17:49:08 +0000 (17:49 +0000)]
Sync usage and man page with reality.  There's no '-c' command line
flag today.  Maybe we should still retain it, but I'll let others fight
that windmill.

19 years agoCleanup handling of colons in variable substitutions.
harti [Fri, 18 Mar 2005 16:52:46 +0000 (16:52 +0000)]
Cleanup handling of colons in variable substitutions.

Patch: 7.126

Submitted by: Max Okumoto <okumoto@ucsd.edu>

19 years agoNeed to initialize the resource list that we keep for our children.
imp [Fri, 18 Mar 2005 16:37:51 +0000 (16:37 +0000)]
Need to initialize the resource list that we keep for our children.
STAILQ's require this, while it is optional for SLIST (well, as long
as the memory is bzeroed).

Noticed by: phk's crash

19 years agoUnfold parsing of single letter modifiers from multi-letter ones.
harti [Fri, 18 Mar 2005 15:51:04 +0000 (15:51 +0000)]
Unfold parsing of single letter modifiers from multi-letter ones.

Patch: 7.125

Submitted by: Max Okumoto <okumoto@ucsd.edu>

19 years agoStyle: fix indentation and <space><tab>.
harti [Fri, 18 Mar 2005 15:44:23 +0000 (15:44 +0000)]
Style: fix indentation and <space><tab>.

Patch: 7.124

Submitted by: Max Okumoto <okumoto@ucsd.edu>

19 years agoSplit out the 'S' modifier into its own function.
harti [Fri, 18 Mar 2005 15:37:31 +0000 (15:37 +0000)]
Split out the 'S' modifier into its own function.

Patch: 7.122,7.123

Submitted by: Max Okumoto <okumoto@ucsd.edu>

19 years agoConvert Var_Parse to a wrapper function.
harti [Fri, 18 Mar 2005 15:34:07 +0000 (15:34 +0000)]
Convert Var_Parse to a wrapper function.
Reduce the number of arguments passed between these functions by
creating a special-purpose struct.

Patch: 7.120,7.121

Submitted by: Max Okumoto <okumoto@ucsd.edu>

19 years agoReplace Lst_Find calls with LST_FOREACH loops. This helps in
harti [Fri, 18 Mar 2005 15:25:23 +0000 (15:25 +0000)]
Replace Lst_Find calls with LST_FOREACH loops. This helps in
constification und simplifies the code because the one-liner
predicates can be inlined into the code.

19 years agoAdd a comment to note that pseudo-device bpf is required for DHCP.
murray [Fri, 18 Mar 2005 15:24:00 +0000 (15:24 +0000)]
Add a comment to note that pseudo-device bpf is required for DHCP.
This is mentioned in the Handbook but it is not as obvious to new
users why bpf is needed compared to the other largely self-explanatory
items in GENERIC.

PR: conf/40855
MFC after: 1 week

19 years agoReplace a bogus use of Lst_Find with explicite LST_FOREACH calls
harti [Fri, 18 Mar 2005 15:23:49 +0000 (15:23 +0000)]
Replace a bogus use of Lst_Find with explicite LST_FOREACH calls
and remove the unneeded second argument to ReadMakefile.

19 years agoFix a bug in matching suffixes. Under certain circumstances the code
harti [Fri, 18 Mar 2005 15:16:09 +0000 (15:16 +0000)]
Fix a bug in matching suffixes. Under certain circumstances the code
would access memory before the beginning of the string to match (the
suffix match starts at the end of both the string and the suffix and
proceedes to the begin until either the start of the suffix is hit
or the character does not match). This could lead to a memcpy copying
into random memory. Fix this by checking the length of the string to
match too and replacing the Lst_Find calls with LST_FOREACH loops
(last part by me).

Submitted by: Matt Dillon <dillon@apollo.backplane.com> (in principle)

19 years agoIf vhid exists return more informative EEXIST instead of EINVAL. While here
glebius [Fri, 18 Mar 2005 13:41:38 +0000 (13:41 +0000)]
If vhid exists return more informative EEXIST instead of EINVAL. While here
remove redundant brackets.

19 years agoImpose the upper limit on signals that are allowed between kernel threads
sobomax [Fri, 18 Mar 2005 13:33:18 +0000 (13:33 +0000)]
Impose the upper limit on signals that are allowed between kernel threads
in set[ug]id program for compatibility with Linux. Linuxthreads uses
4 signals from SIGRTMIN to SIGRTMIN+3.

Pointed out by: rwatson

19 years agoFix a potential crash that could occur when CARP_LOG is being used.
glebius [Fri, 18 Mar 2005 13:18:34 +0000 (13:18 +0000)]
Fix a potential crash that could occur when CARP_LOG is being used.

Obtained from: OpenBSD (pat)

19 years agoRun "make obj" before "make build-tools" in the directories used for
cperciva [Fri, 18 Mar 2005 12:55:07 +0000 (12:55 +0000)]
Run "make obj" before "make build-tools" in the directories used for
building the rescue binary.  This fixes a problem with NO_TCSH, where
the "make obj" stage of buildworld doesn't recurse into bin/csh,
resulting in csh build-tools being put into /usr/src/bin/csh.

Pointed out by: dougb (on hackers@)

19 years agoUse subr_unit to allocate thread ID's with.
phk [Fri, 18 Mar 2005 12:34:14 +0000 (12:34 +0000)]
Use subr_unit to allocate thread ID's with.

Tested by: davidxu

19 years agos/SLIST/STAILQ/
phk [Fri, 18 Mar 2005 11:57:44 +0000 (11:57 +0000)]
s/SLIST/STAILQ/
/imp/a\
pointy hat
.

19 years agoLinuxthreads uses not only signal 32 but several signals >= 32.
sobomax [Fri, 18 Mar 2005 11:08:55 +0000 (11:08 +0000)]
Linuxthreads uses not only signal 32 but several signals >= 32.

PR: kern/72922
Submitted by: Andriy Gapon <avg@icyb.net.ua>

19 years agomakebdev() is long gone.
phk [Fri, 18 Mar 2005 10:57:11 +0000 (10:57 +0000)]
makebdev() is long gone.

19 years ago- autofs was removed some time ago and libautofs is not hooked up to
brueffer [Fri, 18 Mar 2005 10:21:35 +0000 (10:21 +0000)]
- autofs was removed some time ago and libautofs is not hooked up to
  the build.  Remove both from here.
- acpi_snc was renamed to acpi_sony some time ago.

19 years agoNew release notes:
hrs [Fri, 18 Mar 2005 09:47:50 +0000 (09:47 +0000)]
New release notes:
acpi_fujitsu driver added.

19 years agoSplit configure() into 3 separate steps like we do on other
iedowse [Fri, 18 Mar 2005 09:45:43 +0000 (09:45 +0000)]
Split configure() into 3 separate steps like we do on other
architectures. This makes it possible to insert hooks before and
after the device attachment step.

Tested thanks to: marcel

19 years agoHook acpi_fujitsu up to the build.
philip [Fri, 18 Mar 2005 09:34:52 +0000 (09:34 +0000)]
Hook acpi_fujitsu up to the build.

Forgotten by: philip

19 years agoAdd acpi_fujitsu for handling acpi-controlled buttons on Fujitsu laptops.
philip [Fri, 18 Mar 2005 08:48:10 +0000 (08:48 +0000)]
Add acpi_fujitsu for handling acpi-controlled buttons on Fujitsu laptops.

Submitted by: Anish Mistry <mistry.7 -at- osu.edu>
Reviewed by: njl
X-MFC after: 5.4-RELEASE

19 years agoFix off-by-one (too little!) array size problem.
phk [Fri, 18 Mar 2005 07:13:35 +0000 (07:13 +0000)]
Fix off-by-one (too little!) array size problem.

Detected by: Coverity (ID#661)

19 years agog_read_data() can return NULL, check for it.
phk [Fri, 18 Mar 2005 07:03:56 +0000 (07:03 +0000)]
g_read_data() can return NULL, check for it.

Found by: Coverity (ID#258)

19 years agoAfter rejecting the bio request early, return instead of panicing.
phk [Fri, 18 Mar 2005 07:01:31 +0000 (07:01 +0000)]
After rejecting the bio request early, return instead of panicing.

Found by: Coverity (ID#450)

19 years agoAvoid null pointer dereference.
phk [Fri, 18 Mar 2005 06:57:58 +0000 (06:57 +0000)]
Avoid null pointer dereference.

19 years agoFix a bad copy&paste mistake I made.
phk [Fri, 18 Mar 2005 06:01:21 +0000 (06:01 +0000)]
Fix a bad copy&paste mistake I made.

Spotted by: truckman

19 years agoIf mlx_attach() returns an error, don't free sc again.
das [Fri, 18 Mar 2005 05:43:37 +0000 (05:43 +0000)]
If mlx_attach() returns an error, don't free sc again.

Spotted by: Ted Unangst using the Coverity Prevent static analysis tool
Reviewed by: scottl

19 years agoDon't read past the end of pVDevice[]. (Previously, we would iterate
das [Fri, 18 Mar 2005 05:43:34 +0000 (05:43 +0000)]
Don't read past the end of pVDevice[].  (Previously, we would iterate
twice as many times as there were entries in the array.)

Spotted by: Ted Unangst using the Coverity Prevent static analysis tool
Reviewed by: scottl

19 years agoDon't write past the end of the VendorId field (and into the ProductId
das [Fri, 18 Mar 2005 05:43:31 +0000 (05:43 +0000)]
Don't write past the end of the VendorId field (and into the ProductId
field).

Spotted by: Ted Unangst using the Coverity Prevent static analysis tool
Reviewed by: scottl

19 years agoUse STAILQ in preference to SLIST for the resources. Insert new resources
imp [Fri, 18 Mar 2005 05:19:50 +0000 (05:19 +0000)]
Use STAILQ in preference to SLIST for the resources.  Insert new resources
last in the list rather than first.

This makes the resouces print in the 4.x order rather than the 5.x order
(eg fdc0 at 0x3f0-0x3f5,0x3f7 is 4.x, but 0x3f7,0x3f0-0x3f5 is 5.x).  This
also means that the pci code will once again print the resources in BAR
ascending order.

19 years agoSplit configure into 3 steps ala sparc64
grehan [Fri, 18 Mar 2005 03:29:39 +0000 (03:29 +0000)]
Split configure into 3 steps ala sparc64

Obtained from:  iedowse, sparc64

19 years agoAutogenerate device listings for sr(4) and rc(4).
brueffer [Fri, 18 Mar 2005 03:12:47 +0000 (03:12 +0000)]
Autogenerate device listings for sr(4) and rc(4).

MFC after: 3 days

19 years agoFurther refine information about supported hardware (from the hardware notes).
brueffer [Fri, 18 Mar 2005 02:49:43 +0000 (02:49 +0000)]
Further refine information about supported hardware (from the hardware notes).

19 years ago- move supported hardware information into a HARDWARE section
brueffer [Fri, 18 Mar 2005 02:38:03 +0000 (02:38 +0000)]
- move supported hardware information into a HARDWARE section
- consistently capitalize RISCom
- fix typo
- bump .Dd

MFC after: 3 days

19 years agoFix the double rounding problem with subnormals, and
das [Fri, 18 Mar 2005 02:27:59 +0000 (02:27 +0000)]
Fix the double rounding problem with subnormals, and
remove the XXX comments, which no longer apply.

19 years ago- add a HARDWARE section
brueffer [Fri, 18 Mar 2005 02:26:05 +0000 (02:26 +0000)]
- add a HARDWARE section
- correct a language nit
- remove outdated FILES section
- bump .Dd

MFC after: 3 days

19 years agoReflect devnode naming change from /dev/ubser?.? to /dev/ttyy??
ticso [Fri, 18 Mar 2005 02:25:45 +0000 (02:25 +0000)]
Reflect devnode naming change from /dev/ubser?.? to /dev/ttyy??

19 years agoAdd missing prototypes for fma() and fmaf(), and remove an inaccurate
das [Fri, 18 Mar 2005 01:47:42 +0000 (01:47 +0000)]
Add missing prototypes for fma() and fmaf(), and remove an inaccurate
comment.

19 years agofix aio+kq... I've been running ambrisko's test program for much longer
jmg [Fri, 18 Mar 2005 01:11:39 +0000 (01:11 +0000)]
fix aio+kq... I've been running ambrisko's test program for much longer
w/o problems than I was before...  This simply brings back the knote_delete
as knlist_delete which will also drop the knote's, instead of just clearing
the list and seeing _ONESHOT...

Fix a race where if a note was _INFLUX and _DETACHED, it could end up being
modified... whoopse..

MFC after: 1 week
Prodded by: ambrisko and dwhite

19 years agoUse the correct variable name for the description of multi-volume
murray [Fri, 18 Mar 2005 01:05:46 +0000 (01:05 +0000)]
Use the correct variable name for the description of multi-volume
support in cdrom.inf.  Should be CD_VOLUME.

19 years agoRecover gracefully if the user puts in the wrong CD volume after being
murray [Fri, 18 Mar 2005 00:55:12 +0000 (00:55 +0000)]
Recover gracefully if the user puts in the wrong CD volume after being
prompted to insert another CD for a package.

MFC After: 3 days

19 years agoBring back some of the cleanups and fixes jmg did in the TS7200 port.
cognet [Thu, 17 Mar 2005 23:01:15 +0000 (23:01 +0000)]
Bring back some of the cleanups and fixes jmg did in the TS7200 port.

19 years agoinitialize pp->p_sc so it can be referenced later.
ticso [Thu, 17 Mar 2005 22:47:18 +0000 (22:47 +0000)]
initialize pp->p_sc so it can be referenced later.
dynamicaly allocate the per port array.
allow up to 32 serials per USB device.
ask the device for correct pipe sizes.

19 years agoIntroduce a general name for the previously cmbat-only ioctls. It has the
njl [Thu, 17 Mar 2005 22:42:49 +0000 (22:42 +0000)]
Introduce a general name for the previously cmbat-only ioctls.  It has the
same value as the previous ioctls so no binary change.  Also, make a few
style changes to reduce diffs to my tree.

Loosely based on code from: Hans Petter Selasky

19 years agoRemove the 'usbd' keyword (it isn't necessary for mixer). Also, use
njl [Thu, 17 Mar 2005 22:36:16 +0000 (22:36 +0000)]
Remove the 'usbd' keyword (it isn't necessary for mixer).  Also, use
BEFORE instead of REQUIRE.

Probably ok by: jhb
MFC after: 3 days

19 years agoMake the fenv.h routines work for programs that use SSE for
das [Thu, 17 Mar 2005 22:21:46 +0000 (22:21 +0000)]
Make the fenv.h routines work for programs that use SSE for
floating-point arithmetic on i386.  Now I'm going to make excuses
for why this code is kinda scary:

- To avoid breaking the ABI with 5.3-RELEASE, we can't change
  sizeof(fenv_t).  I stuck the saved mxcsr in some discontiguous
  reserved bits in the existing structure.

- Attempting to access the mxcsr on older processors results
  in an illegal instruction exception, so support for SSE must
  be detected at runtime.  (The extra baggage is optimized away
  if either the application or libm is compiled with -msse{,2}.)

I didn't run tests to ensure that this doesn't SIGILL on older 486's
lacking the cpuid instruction or on other processors lacking SSE.
Results from running the fenv regression test on these processors
would be appreciated.  (You'll need to compile the test with
-DNO_STRICT_DFL_ENV.)  If you have an 80386, or if your processor
supports SSE but the kernel didn't enable it, then you're probably out
of luck.

Also, I un-inlined some of the functions that grew larger as a result
of this change, moving them from fenv.h to fenv.c.

19 years agoInitialize the mxcsr properly, so the initial value in a process isn't
das [Thu, 17 Mar 2005 22:21:36 +0000 (22:21 +0000)]
Initialize the mxcsr properly, so the initial value in a process isn't
just the value that was left over from some other application.

19 years agoFix typo in comments (spell Linux correctly)
emax [Thu, 17 Mar 2005 21:39:44 +0000 (21:39 +0000)]
Fix typo in comments (spell Linux correctly)

Submitted by: Markus Brueffer < markus at brueffer dot de >
MFC after: 3 days

19 years agoWhen locking a MTX_SPIN, one needs to use mtx_lock_spin.
imp [Thu, 17 Mar 2005 20:45:24 +0000 (20:45 +0000)]
When locking a MTX_SPIN, one needs to use mtx_lock_spin.
Lock the timeout routine as well.

Submitted by: bde

19 years agoSplit configure() into 3 separate steps like we do on other
iedowse [Thu, 17 Mar 2005 20:31:36 +0000 (20:31 +0000)]
Split configure() into 3 separate steps like we do on other
architectures. This makes it possible to insert hooks before and
after the device attachment step.

19 years agoNew release notes: BIND 9.3.1.
hrs [Thu, 17 Mar 2005 19:55:07 +0000 (19:55 +0000)]
New release notes: BIND 9.3.1.

19 years agoI keep forgetting that I wanted the customize logs collected too.
phk [Thu, 17 Mar 2005 19:45:05 +0000 (19:45 +0000)]
I keep forgetting that I wanted the customize logs collected too.

Reminded by: Lennart Sorth, Andrea Campi and others.

19 years agoDefer boot-time exploration of USB busses until all devices in the
iedowse [Thu, 17 Mar 2005 19:41:19 +0000 (19:41 +0000)]
Defer boot-time exploration of USB busses until all devices in the
system have been attached, but no later. This ensures that we do
not explore ohci or uhci busses before the companion echi controller
has been initialised, so it should fix the problem of multi-speed
USB devices getting attached as USB 1 devices first and then
re-attached as USB 2.

Some further changes are needed on architectures that do not currently
allow hooks to be inserted before configure_final() - alpha, ia64,
powerpc and sparc64. On these architectures the exploration will
now be delayed until the usb kthread runs.

19 years agoadd m_copyup function.. This can be used to help make our ip stack less
jmg [Thu, 17 Mar 2005 19:34:57 +0000 (19:34 +0000)]
add m_copyup function..  This can be used to help make our ip stack less
alignment restrictive, and help performance on some ethernet cards which
currently copy the entire packet a couple bytes to get the packet aligned
properly...

Wordsmithing by: dwhite
Obtained from: NetBSD (code only)
I'll clean it up later: rwatson

19 years agoeisa attachment is safe to be in this module, both on eisa and
imp [Thu, 17 Mar 2005 18:40:37 +0000 (18:40 +0000)]
eisa attachment is safe to be in this module, both on eisa and
non-eisa configured kernels.

19 years agoNow that the Adaptec 2842 has its own probe routine, no need to have
imp [Thu, 17 Mar 2005 17:36:07 +0000 (17:36 +0000)]
Now that the Adaptec 2842 has its own probe routine, no need to have
a comment saying its probe routine needs to be fixed.

19 years agoStyle(9) pass before some planned larger changes.
imp [Thu, 17 Mar 2005 17:33:52 +0000 (17:33 +0000)]
Style(9) pass before some planned larger changes.
o return (foo);
o if (a == NULL) in preference to if (!a)
o () and {} reduction
o minor indentation fixes

19 years agoA further step on the journey of meaking panics and debugging more reliable:
rwatson [Thu, 17 Mar 2005 15:18:01 +0000 (15:18 +0000)]
A further step on the journey of meaking panics and debugging more reliable:
in the window between the beginning of panic() and entering the debugger,
it's possible to receive interrupts.  If we receive an interrupt, don't
preempt if panicstr != NULL, as the system is in the process of failing, and
the preempting thread is likely to stumble over the failure.  The typical
scenario is during the printf() in panic() prior to entering the debugger,
but when running with a slower console type such as serial console.

It could be that the panic string should be passed to the debugger to print,
so that it can run from the debugger's environment rather than a regular
kernel printf.

Glanced at by: jhb

19 years agoAlso remember to set the fsid here.
phk [Thu, 17 Mar 2005 15:15:29 +0000 (15:15 +0000)]
Also remember to set the fsid here.

19 years agoForgot to replace code to set fsid in vop_getattr.
phk [Thu, 17 Mar 2005 14:43:40 +0000 (14:43 +0000)]
Forgot to replace code to set fsid in vop_getattr.

19 years agoDo not try to free non allocated memory in error case.
bz [Thu, 17 Mar 2005 14:27:22 +0000 (14:27 +0000)]
Do not try to free non allocated memory in error case.
Do our best to plug some memory leaks (VPD data, jumbo memory buffer,...).
Log if we cannot free because memory still in use[1].
Change locking to avoid ''acquiring duplicate lock of same
  type: "network driver"'' and potential deadlock. Also seems to fix LOR #063.

[1] This change does not solve problems if buffers are still in use when
    unloading if_sk.ko. There is ongoing work which will address jumbogram
    allocations in a more general way.

PR: kern/75677 (with changes, no mii fixes in here)
Tested by: net, Antoine Brodin (slightly different version)
Approved by: rwatson (mentor)
MFC after: 5 days

19 years ago* Lower interrupt moderation timer 200->100.
bz [Thu, 17 Mar 2005 14:23:13 +0000 (14:23 +0000)]
* Lower interrupt moderation timer 200->100.
Obtained from: NetBSD if_sk.c rev. 1.11

* Make interrupt moderation configurable via sysctl/tuneable.
PR: kern/41220
Approved by: rwatson (mentor)

19 years ago* Improve chip identification.
bz [Thu, 17 Mar 2005 14:21:51 +0000 (14:21 +0000)]
* Improve chip identification.
Obtained from: NetBSD if_sk.c rev. 1.11

* Take PHY out of reset for Yukon Lite Rev. A3.
Submitted by: postings on net@ in thread "skc0: no PHY found", 2005-02-22

Tested by: net
Approved by: rwatson (mentor)
MFC after: 5 days

19 years ago* When adding/deleting multicast addresses, only whack the address filter
bz [Thu, 17 Mar 2005 14:18:58 +0000 (14:18 +0000)]
* When adding/deleting multicast addresses, only whack the address filter
  if the interface is marked RUNNING.
Obtained from: NetBSD if_sk.c rev. 1.12

* Don't initialize the card (and start an autonegotiation) every time the IP
  address changes. Makes 'dhclient sk0' invocations way faster and more
  consistant. i.e. one DHCPREQUEST elicits the DHCPACK.
Obtained from: OpenBSD if_sk.c rev. 1.56

* Additional locking changes in sk_ioctl.

PR: kern/61296 should see improvements by the last two.
Approved by: rwatson (mentor)
MFC after: 5 days

19 years agoReleasing TX/RX descriptor dmamaps during device detachment instead of
avatar [Thu, 17 Mar 2005 13:59:30 +0000 (13:59 +0000)]
Releasing TX/RX descriptor dmamaps during device detachment instead of
doing that in bfe_stop().

This should fix a panic recently reported on -current occuring when taking
device down then up.  In the original implementation, an "ifconfig bfe0 down"
triggers bfe_stop(), which also destroys all TX/RX descriptor dmamaps. Hence
the subsequent "ifconfig bfe0 up" would force the device to use those
already-released dmamap and thus panic the kernel.

PR: kern/77804
Submitted by: Frank Mayhar <frank at exit dot com>
Reviewed by: dmlb, sam (mentor)
Tested by: Phil <pcasidy at casidy dot com>, myself
MFC after: 1 week

19 years agoKill MAJOR_AUTO
phk [Thu, 17 Mar 2005 13:37:28 +0000 (13:37 +0000)]
Kill MAJOR_AUTO

19 years agoPrepare for the final onslaught on devices:
phk [Thu, 17 Mar 2005 12:07:00 +0000 (12:07 +0000)]
Prepare for the final onslaught on devices:

Move uid/gid/mode from cdev to cdevsw.

Add kind field to use for devd(8) later.

Bump both D_VERSION and __FreeBSD_version

19 years ago - Don't lock the vnode interlock in vm_object_set_writeable_dirty() if
jeff [Thu, 17 Mar 2005 12:03:42 +0000 (12:03 +0000)]
 - Don't lock the vnode interlock in vm_object_set_writeable_dirty() if
   we've already set the object flags.

Reviewed by: alc

19 years ago - Lock the clearing of v_data so it is safe to inspect it with the
jeff [Thu, 17 Mar 2005 12:00:05 +0000 (12:00 +0000)]
 - Lock the clearing of v_data so it is safe to inspect it with the
   interlock.

Sponsored by: Isilon Systems, Inc.

19 years ago - Lock the clearing of v_data in ufs_reclaim() to prevent a pagefault
jeff [Thu, 17 Mar 2005 11:58:43 +0000 (11:58 +0000)]
 - Lock the clearing of v_data in ufs_reclaim() to prevent a pagefault
   in ffs_lock() when it acesses v_data without the vnlock.

Sponsored by: Isilon Systems, Inc.

19 years agoThere is no longer a KNOWN_DEFECTS file, so don't try to install it.
dougb [Thu, 17 Mar 2005 10:15:29 +0000 (10:15 +0000)]
There is no longer a KNOWN_DEFECTS file, so don't try to install it.

19 years agoRemove unused variable.
phk [Thu, 17 Mar 2005 09:32:37 +0000 (09:32 +0000)]
Remove unused variable.

Detected by: Coverity (ID#704)